{"690589":{"#nid":"690589","#data":{"type":"news","title":"Ph.D. Student Gets the Assist as Bike Robot Performs First Front Flip","body":[{"value":"\u003Cp\u003EA bicycle robot from the Robotics and AI Institute (RAI) in Cambridge, Mass., has become the first to perform an unassisted acrobatic front flip.\u003C\/p\u003E\u003Cp\u003ERAI calls the bicycle robot an ultra-mobility vehicle (UMV). It can reach a height of 3 feet and can jump from the floor onto a platform.\u003C\/p\u003E\u003Cp\u003EThe contributions of a Georgia Tech Ph.D. student helped make these feats possible through a robot control policy he developed.\u003C\/p\u003E\u003Cp\u003EJeonghwan Kim, who is pursuing a Ph.D. in robotics under the advisement of Associate Professor Sehoon Ha, spent two semesters interning at RAI. His task was to design a policy to teach the UMV to land after a flip.\u003C\/p\u003E\u003Cp\u003EThe result was iterative motion imitation (IMI), a novel method that imitates flip trajectories generated from prior examples. Kim said the robot bases its flip on a demonstration, and human engineers reconstruct and refine the flip path through simulation to fill in the gaps.\u003C\/p\u003E\u003Cp\u003E\u201cTo guide the robot to flip, we started with an imperfect trajectory generated by a motor-based controller and then ran simulations,\u201d Kim said. \u201cIt\u2019s an unstable trajectory, but we use it as a guide to train a single policy that can track it as it lands and tries to balance itself.\u201d\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Ch4\u003E\u003Cstrong\u003ESticking the Landing\u003C\/strong\u003E\u003C\/h4\u003E\u003Cp\u003EKim interned under the supervision of Shamel Fahmi, a research scientist at the RAI Institute. RAI has been developing the UMV for nearly three years.\u003C\/p\u003E\u003Cp\u003E\u201cWe wanted to work on a different robot morphology that wasn\u2019t legs or legs with wheels,\u201d Fahmi said. \u201cThat\u2019s when we thought of working with bikes.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u201cWe want to merge the athleticism of (Boston Dynamics\u2019) Atlas with the mobility of a bike. We wanted a robot that can go anywhere, do parkour, and acrobatics.\u201d\u003C\/p\u003E\u003Cp\u003EFahmi said that before Kim arrived, the research team had trouble getting the UMV to land consistently without breaking or falling.\u003C\/p\u003E\u003Cp\u003EThe UMV has two joints \u2014 an upper and a lower. The upper joint contains the motors and pulls the lower joint along as it propels into the air. The problem is getting the lighter lower joint to absorb the impact of landing without being crushed by the heavier upper joint.\u003C\/p\u003E\u003Cp\u003E\u201cThat\u2019s what brings reinforcement learning into the equation,\u201d Fahmi said. \u201cWe teach the robot to minimize its impact on the ground to land gracefully.\u201d\u003C\/p\u003E\u003Cp\u003EFahmi said that Kim proved the imitation examples the robot learns from don\u2019t have to be perfect. The process takes some time, but all it needs is a rough idea to get started.\u003C\/p\u003E\u003Cp\u003E\u201cYou can have an imperfect sketch and then constantly refine it,\u201d Fahmi said. \u201cThe first time, it\u2019s not going to go well.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u201cWe don\u2019t care about torque or power limits as long as it does the motion. Then we\u2019ll have a slightly better reference, repeat it, and imitate it again. In every iteration, we can add more parameters.\u201d\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Ch4\u003E\u003Cstrong\u003EUp Against the Clock\u003C\/strong\u003E\u003C\/h4\u003E\u003Cp\u003EKim said he felt the pressure of time constraints during his two semesters with RAI as he worked to achieve consistent, successful landings. Even though he had multiple UMVs to experiment with, they broke down dozens of times. Each time one broke, a hardware team at RAI had to repair it.\u003C\/p\u003E\u003Cp\u003E\u201cThere was a lot of pressure to not only get this working before my internship ended, but also knowing there are costs behind every failed attempt, and every time the robot breaks, it takes time to repair it,\u201d Kim said.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u201cIt took almost five months for it to land without breaking. Then we needed two more months for it to stay balanced after the landing. It requires a lot of engineering effort to achieve a robust control policy for a safe flip.\u201d\u003C\/p\u003E\u003Cp\u003EBy the time Kim left RAI, the IMI policy had achieved consistent, seamless landings.\u003C\/p\u003E\u003Cp\u003E\u201cThe jump right now is what we call the visitor demo,\u201d Fahmi said. \u201cIf there are guests coming over to see it, we want to show them something that is extremely impressive, but also, more importantly, extremely reliable. It never fails.\u003C\/p\u003E\u003Cp\u003E\u201cIt was only possible because of the huge effort we put into designing, maintaining, and continuously improving the robot.\u201d\u003C\/p\u003E\u003Cp\u003EKim authored a\u0026nbsp;\u003Ca href=\u0022https:\/\/imi-umv.github.io\/\u0022\u003Epaper\u003C\/a\u003E on his framework and will present it at this week\u2019s\u0026nbsp;\u003Ca href=\u0022https:\/\/2026.ieee-icra.org\/\u0022\u003EInternational Conference on Robotics and Automation\u003C\/a\u003E (ICRA) in Vienna.\u003C\/p\u003E\u003Cp\u003EFor more information about the UMV project, please visit the\u0026nbsp;\u003Ca href=\u0022https:\/\/rai-inst.com\/resources\/blog\/designing-wheeled-robotic-systems\/\u0022\u003ERAI blog\u003C\/a\u003E or watch their\u0026nbsp;\u003Ca href=\u0022https:\/\/www.youtube.com\/watch?v=cjaZUFMZWOY\u0026amp;t=95s\u0022\u003Evideo\u003C\/a\u003E on YouTube.\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EA bicycle robot from the Robotics and AI Institute (RAI) in Cambridge, Mass., has become the first to perform an unassisted acrobatic front flip.\u003C\/p\u003E\u003Cp\u003EJeonghwan Kim, who is pursuing a Ph.D. in robotics under the advisement of Associate Professor Sehoon Ha, spent two semesters interning at RAI. His task was to design a policy to teach the UMV to land after a flip.\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"A Georgia Tech Ph.D. student\u0027s robot control policy helped the Robotics and AI Institute develop the first bike robot capable of an unassisted front flip."}],"uid":"36530","created_gmt":"2026-06-02 13:06:16","changed_gmt":"2026-06-02 13:11:56","author":"Nathan Deen","boilerplate_text":"","field_publication":"","field_article_url":"","location":"Atlanta, GA","dateline":{"date":"2026-06-02T00:00:00-04:00","iso_date":"2026-06-02T00:00:00-04:00","tz":"America\/New_York"},"extras":[],"hg_media":{"680398":{"id":"680398","type":"image","title":"DSC07117-2048x1365.jpg-copy.jpg","body":"\u003Cp\u003EPhoto courtesy of the Robotics and AI Institute\u003C\/p\u003E","created":"1780405593","gmt_created":"2026-06-02 13:06:33","changed":"1780405662","gmt_changed":"2026-06-02 13:07:42","alt":"Bike robot","file":{"fid":"264654","name":"DSC07117-2048x1365.jpg-copy.jpg","image_path":"\/sites\/default\/files\/2026\/06\/02\/DSC07117-2048x1365.jpg-copy.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2026\/06\/02\/DSC07117-2048x1365.jpg-copy.jpg","mime":"image\/jpeg","size":131530,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2026\/06\/02\/DSC07117-2048x1365.jpg-copy.jpg?itok=12RGh4JN"}}},"media_ids":["680398"],"groups":[{"id":"47223","name":"College of Computing"},{"id":"1188","name":"Research Horizons"},{"id":"50876","name":"School of Interactive Computing"}],"categories":[{"id":"152","name":"Robotics"}],"keywords":[{"id":"188776","name":"go-research"},{"id":"187915","name":"go-researchnews"},{"id":"9153","name":"Research Horizons"},{"id":"187991","name":"go-robotics"},{"id":"184632","name":"mobile robotics"}],"core_research_areas":[{"id":"39521","name":"Robotics"}],"news_room_topics":[],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}